AlphaGo Zero: Google DeepMind supercomputer learns 3,000 years of human knowledge in 40 days

Discussion in 'Off Topic' started by Dr. AMK, Oct 20, 2017.

  1. Dr. AMK

    Dr. AMK Notebook Evangelist

    Reputations:
    470
    Messages:
    494
    Likes Received:
    945
    Trophy Points:
    106
    AlphaGo Zero: Google DeepMind supercomputer learns 3,000 years of human knowledge in 40 days
    [​IMG]
    Chinese Go player Ke Jie competes against Google's artificial intelligence (AI) program, AlphaGoCREDIT: IMAGINECHINA/REX/SHUTTERSTOCK

    Thousands of years of human knowledge has been learned and surpassed by the world’s smartest computer in just 40 days, a breakthrough hailed as one of the greatest advances ever in artificial intelligence.

    Google DeepMind amazed the world last year when its AI programme AlphaGo beat world champion Lee Sedol at Go, an ancient and complex game of strategy and intuition which many believed could never be cracked by a machine.

    AlphaGo was so effective because it had been programmed with millions of moves of past masters, and could predict its own chances of winning, adjusting its game-plan accordingly.

    But now the same team has created a machine that learns from scratch.
     
    hmscott likes this.
  2. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,627
    Messages:
    13,487
    Likes Received:
    15,609
    Trophy Points:
    931
    Wow, I thought we had progressed to the point in societal general computer functional understanding where promoting such outlandish hyperbole concerning computers "learning and thinking" was not possible, and certainly not as promulgated by a company setting the societal standards such as Google.

    What a load of male bovine excrement. :)

    It's a program that plays go. It's designed to process and store positive and negative outcome branches, and to rapidly search them for solutions. I haven't gone into the totality of it's functional device modules, there are lots of tricks to improve performance. But the idea of it all hasn't changed since I wrote my first go program in the 70's.

    HUman Learning is a much different process, and encompasses a wide range of all experienced outcomes of all parts of our lives. We are not as humans "go programs", we are go players, while simultaneously being infinitely more than that.

    It didn't "Learn 3,000 years of human knowledge", that's an extremely loose and open definition of what it did / does, and is the crux of my complaint on their promotion of their accomplishment. It's total fabrication and ludicrously false.

    The mechanics of elements of the handling of storing response from action may be applicable to other endeavors, but as written it's a Go program, it plays go, that's it.
     
    Last edited: Oct 27, 2017
    Kyle and Dr. AMK like this.
  3. Dr. AMK

    Dr. AMK Notebook Evangelist

    Reputations:
    470
    Messages:
    494
    Likes Received:
    945
    Trophy Points:
    106
    hmscott likes this.
  4. Dr. AMK

    Dr. AMK Notebook Evangelist

    Reputations:
    470
    Messages:
    494
    Likes Received:
    945
    Trophy Points:
    106
    I agree with you, and I hope that they didn't go that far. I don't like any AI to exceed any of the red lines, it too dangerous.
     
    Last edited: Oct 20, 2017
    hmscott likes this.
  5. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,627
    Messages:
    13,487
    Likes Received:
    15,609
    Trophy Points:
    931
    It's a worthy accomplishment in the progression of Knowledge-based Expert Systems, which has been ongoing for a long time.

    Singling out a new technique as "learning" is a great disservice, and leads to uninformed experts in other fields laughingly cowering at "AI Overloads" defeating mankind in the near future.

    I love SciFi Fantasy films too, but we are a long way from living inside of one... ;)
     
    Dr. AMK likes this.
  6. Dr. AMK

    Dr. AMK Notebook Evangelist

    Reputations:
    470
    Messages:
    494
    Likes Received:
    945
    Trophy Points:
    106
    sci-fi-o-rama_uibaka_011.gif
    :)
     
    hmscott likes this.
  7. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,627
    Messages:
    13,487
    Likes Received:
    15,609
    Trophy Points:
    931
    Dr. AMK likes this.
  8. Jarhead

    Jarhead Systematic Love

    Reputations:
    4,816
    Messages:
    11,863
    Likes Received:
    2,470
    Trophy Points:
    581
    Just to be clear, does the “3,000 years of human knowledge” bit in the title refer to all human knowledge, or all human knowledge pertaining to Go? It’s a bit ambiguous based on the contents of the article.

    Anyway, the Chinese Room and philosophy of mind come to mind here. You can also say that human intelligence is also simply just electrical impulses between nodes in a system (since that is what the brain is, after all), so I wouldn’t be too quick to judge about what is “real” and “fake” intelligence. The real question is if a computer can become self-aware or have some sort of self-determination.
     
    Dr. AMK and hmscott like this.
  9. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    3,627
    Messages:
    13,487
    Likes Received:
    15,609
    Trophy Points:
    931
    Yes, you caught them out exactly, it's Go Knowledge.

    In fact if their game is AI intelligence, via inference learning engine, then that 3,000 years of Knowledge wasn't transfered as implied, but rather inferred by beating some poor schmuck 100 games out of 100 games.

    I don't think that losing Go guy had 3,000 years of Knowledge either ;)
     
    Fishon and Dr. AMK like this.
  10. Jarhead

    Jarhead Systematic Love

    Reputations:
    4,816
    Messages:
    11,863
    Likes Received:
    2,470
    Trophy Points:
    581
    Well, by that sort of statement, nobody has knowledge of Go since I find it hard to believe that any player, human or computer, has a perfect Go record. Still, I wouldn’t conclude that those regarded as Go masters have no Go knowledge.

    It’s certainly possible for a human to study up on 3,000 years of Go knowledge (hell, that data had to come from somewhere), but the main advantage of a computer is that computers have “instant” memory recall (as least insofar as cache/RAM/SSD/HDD/etc performance goes) and can perform calculations of game trees a lot faster than a human could. They’re already much better at a lot of things than we are, with human’s main advantage over computers being pattern recognition (a computer will have a hard time seeing a face in the clouds, for example) and non-intelligent actions (such as conveying emotions, useful for certain service jobs).

    That all said, as a computer scientist I’m not too happy to see a media outlet intentionally being misleading about their articles/titles in order to gain clicks. Gotta love journalism.
     
    Dr. AMK and hmscott like this.
Loading...

Share This Page